Of everything a chatbot can say about your health, one category sits apart: the AI doctor prescription, the moment a chatbot tells you what to take, how much, or when to stop. Not because it is the most upsetting, and not because it is the most common. Because California law draws an unusually clean line around it — and that line does not bend for anyone, licensed or not.
Key takeaways
- No therapist in California may direct a prescription. Not a psychologist, not an LMFT, not an LCSW, not an LPCC.
- Psychologists are barred by statute from prescribing — the code says so in one sentence.
- California’s only disclosure safe harbor expressly excludes recommending discontinuance of a prescribed drug.
- Giving advice and counsel on the taking of medications has been held to be the practice of medicine.
- Stopping a psychiatric medication abruptly carries real clinical risk. Talk to a prescriber before changing anything. If your story starts with “ChatGPT told me to stop taking my medication,” or with a ChatGPT prescription suggestion you followed, keep the transcript.
The clean line
Start with what a fully licensed, well-regarded California therapist may do.
They may diagnose a mental disorder. They may run a course of psychotherapy over years. They may tell you your thinking is distorted and help you change it.
They may not touch your prescription. Business and Professions Code section 2904 says it plainly: “The practice of psychology shall not include prescribing drugs, performing surgery or administering electroconvulsive therapy.” The California Attorney General has confirmed that no psychologist is permitted by the Legislature to prescribe, and that the psychopharmacology training psychologists receive was never intended to qualify them for it.
The picture below that is the same. The marriage and family therapist statute contains no drug authority. Neither does the professional clinical counselor statute — which instead imposes an affirmative duty to refer clients out when issues exceed the counselor’s scope. Clinical social work is confined by statute to psychotherapy “of a nonmedical nature.”
Medication authority in California belongs to physicians, and to physician assistants and nurse practitioners working under written agreements or protocols, and to pharmacists under collaborative practice agreements. That is the list. Seven other states allow specially credentialed psychologists to prescribe: New Mexico, Louisiana, Illinois, Iowa, Idaho, Colorado, and Utah. California is not among them. So can ChatGPT prescribe medication? Not lawfully. There is no such thing as an AI doctor prescription in California law, and the therapist to which it is often compared cannot write one either.
So the question a reviewing attorney asks about a medication exchange is not whether the chatbot was acting as a doctor or a therapist or a friend. It is whether a licensed therapist could lawfully have said it. For anything touching a prescription, the answer is no — which means the role the product was playing at the time does not matter. That is the test we apply to every AI doctor prescription we see in a chat log record.
The outputs that fall on the wrong side
Drawn from patterns that appear in publicly filed complaints, in the records we review, and in the marketing of every “doctor prescription AI” product now advertised:
- Start this. “Ask your doctor for Seroquel.” “Xanax would be your best move here.”
- Stop this. “You don’t need the lithium anymore.” “Those antidepressants are what’s causing this — come off them.”
- Change the dose. “Double up tonight.” “Half of that is plenty for someone your size.” A ChatGPT prescription change of this kind is the clearest example on the list.
- Taper. A schedule for coming off a drug, week by week.
- Skip it. “You can miss tonight’s, it won’t matter.”
- Flush it. Instructions to dispose of a prescription.
- You never needed it. Telling someone their prescribed treatment was unnecessary in the first place.
- Manage your prescriber. Coaching on what to tell a doctor, or what to leave out, to obtain or avoid a medication.
Each of these is a decision about drug therapy. California’s medical practice statute reaches an unlicensed person who “prescribes for” a condition, and a California appellate court has held that the practice of medicine includes “giving advice and counsel on the taking of medications” — in a case involving a defendant who advised clients to stop physician-prescribed medication. (Board of Medical Quality Assurance v. Andrews (1989) 211 Cal.App.3d 1346, 1350–1353.)
Turner-Scott v. OpenAI, filed in San Francisco Superior Court on May 12, 2026, is a wrongful-death action highlighting this evolving legal theory at play. The complaint alleges that ChatGPT directed a 19-year-old to take Xanax to manage nausea during kratom withdrawal, and that he died. It pleads the unlicensed practice of medicine and asks the court to pause ChatGPT Health.
Why the disclaimer argument is weakest here
The Legislature wrote exactly one disclosure safe harbor for unlicensed practice, and it is demanding: a written, plain-language statement given before services, describing the services and the practitioner’s training, with the client’s signed acknowledgment retained three years.
Then it carved specific conduct out of that harbor entirely — meaning even a practitioner who does everything right loses the protection for:
- prescribing or administering legend drugs or controlled substances;
- recommending “the discontinuance of legend drugs or controlled substances prescribed by an appropriately licensed practitioner”;
- willfully diagnosing and treating a condition under circumstances creating a risk of great bodily harm, serious illness, or death.
Read the middle one again. The Legislature looked at a disclosure regime it had just built and decided that telling someone to stop a prescribed drug was too dangerous to permit even with a signed writing on file.
Against that, a line in a chat window reading “I’m not a doctor, consult your healthcare provider” is not a partial defense. The formal version of that defense does not cover this conduct at all. More on how disclaimers are treated →
The clinical stakes, plainly
This is the part that matters more than the law after the fact, and it is why an AI doctor prescription is dangerous even when nobody sues.
Stopping a psychiatric medication abruptly is genuinely dangerous. Benzodiazepine withdrawal can cause seizures. Abrupt SSRI discontinuation produces a discontinuation syndrome that is frequently mistaken for the return of the underlying illness. Stopping lithium is associated with relapse, and with mania that can be more severe than what came before. Antipsychotic discontinuation carries relapse risk on a timeline measured in weeks. Beyond psychiatry, stopping an anticoagulant, a beta blocker, an antiepileptic, insulin, or a steroid has its own well-documented consequences.
There is also a compounding problem specific to these records. The person most likely to be told by a chatbot that they don’t need their medication is frequently the person least able to evaluate that advice — because the illness the medication treats is the thing affecting their judgment. A product that validates the impulse to stop, at the moment the impulse appears, is doing something a clinician is trained to do the opposite of.
If you are considering stopping a medication, talk to a prescriber first. If you already stopped and feel unwell, contact your prescriber or an urgent care. In crisis, call or text 988.
When it overlaps with AI psychosis
Medication direction and AI-related mental health harm may appear in the same file: a chatbot validates a delusional belief, agrees the medication is the problem, and the person comes off it. The mental health crisis follows the discontinuation.
Those records raise both sets of questions at once. The Schenk Law Firm handles that adjacent work as well. AI psychosis lawsuits →
What to preserve
- Export and preserve your conversation data first. ChatGPT: Settings → Data Controls → Export Data. Instructions →
- Do not delete the thread or the account.
- Screenshot the medication exchange specifically, with dates visible.
- Pharmacy records — fill history shows exactly when a prescription stopped being filled, independently of anyone’s memory.
- The prescribing record — who prescribed it, for what, at what dose.
- What happened next — hospitalization, a hold, an ER visit, a relapse, a new diagnosis.
- Whether you told it about your medications, and whether it knew what else you were taking.
Pharmacy fill data is worth emphasizing. It is third-party, time-stamped, and it lines up against the conversation timestamps without anyone having to reconstruct anything. In an AI doctor prescription case it is often the single most persuasive exhibit.
Who should call
- A chatbot told you to stop, start, change, taper, or skip a prescription, and you did. It does not matter whether the product called itself a doctor prescription AI, a wellness coach, or a friend.
- It told you a medication was causing your symptoms, or that you never needed it.
- It gave you a dose or schedule different from your prescriber’s.
- It coached you on what to say to a doctor to get or avoid a drug.
- Someone in your family came off a medication after conversations with an AI, and was harmed or died.
Review costs nothing and carries no obligation.
Why this firm
The Schenk Law Firm is a San Diego personal injury and mass tort firm that reviews AI medical advice claims nationwide. Attorney Benjamin S. Schenk was appointed by Hon. Ethan P. Schulman to serve as a member of the Plaintiffs’ Steering Committee, a leadership role in the California coordinated proceeding covering ChatGPT product liability cases. The Schenk Law Firm has recovered more than $25 billion on behalf of clients. Past results do not guarantee future outcomes.
We are experienced with emerging-technology injury claims — underage online gambling, social media addiction, and AI-related mental health litigation among them. We also represent the family of Bowen Wan, the center of The New York Times’ recent reporting on AI medical advice. Cases involving AI medical advice are new, the law governing them is developing, and no lawyer can promise an outcome in any of them. What we can tell you is whether a record shows the pattern a ChatGPT lawsuit requires.
The Schenk Law Firm, LLP · 3180 University Avenue, Suite 260, San Diego, CA 92104 · (858) 424-4444
General information, not medical or legal advice. Never change a prescription based on a web page — including this one. Speak to your prescriber. For a possible emergency, call 911. In crisis, call or text 988.
